The primary goal of this study is to develop an automated scoring system of leptomeningeal collaterals based on CT angiography (CTA) as a quantitative assessment of acute ischemic stroke. Leptomeningeal collaterals provide an alternate path of blood flow in the setting of a proximal vessel occlusion. The presence and effectiveness of leptomeningeal collaterals (i.e., ?collateral status?) varies significantly from patient to patient. Collateral status has been shown to be correlated with both patient outcome and risk of hemorrhagic transformation, suggesting that rapid and accurate assessment of the collaterals would provide a powerful tool for treatment evaluation. The current gold standard for collateral scoring is performed by subjective analysis of retrograde filling in invasive digital subtraction angiograms (DSA). Though CTA is more limited than DSA in the direct evaluation of collaterals, the presence of collaterals may also be evaluated in CTA by the tree-pattern of vessel filling around the site of occlusion. Several manual assessment techniques based on CTA have been proposed in the literature, but all are subjective and require significant human intervention. During Phase I, we deployed a web-based system for automated assessment of stroke collaterals based on CTA, and we used it in a preliminary study of collateral status for stroke assessment. In this Phase II STTR proposal, the specific aims are (1) to improve the integration of our system with clinical workflows; (2) expand the system to also compute CT perfusion stroke measures, so as to provide a unified platform for comparing and combining CTA collateral and CT perfusion assessment metrics; and (3) test the hypotheses that the combined use of CTA collateral and CT perfusion measures outperforms either method independently for treatment efficacy prediction in stroke patients.
Leptomeningeal collateral vessels provide connections between vascular territories in the brain when occlusions occur, e.g., when a person suffers a stroke. The number and connectivity of collateral vessels varies significantly from patient to patient, and they have been shown to play an important role in stroke outcomes, providing blood flow to tissue at risk. Rapid and effective patient-specific determination of the presence and connectivity of collaterals is needed in acute stroke situations to determine which treatments will be effective. We propose to develop an automated system to evaluate collateral vessel trees from computed tomography angiography (CTA) images and to offer that system as an algorithms-as-a-service (AaaS) product.
